Changes to imputation of INCWAGE

I would like to identify the percent of the CPS-ASEC with imputed INCWAGE from 1968 onwards.

To achieve this, I tried using a combination of QINCWAGE, QOINCWAGE, and QINCLONGJ. I used QINCWAGE for observations in 1968-87 and QOINCWAGE and INCLONGJ (if SRCEARN = 1) for observations 1988 and later. Is this the correct way look at the imputation of INCWAGE over time?

Using the above method yields a pretty large jump in 1988 in the percent of observations that are imputed. Among my sample (single women) it goes from roughly 13% imputations from 1968-87 to 6% in 1988.

You are correct. The flag variable specifically for INCWAGE (e.g., QINCWAGE) is only available in 1962 and from 1968-1987. In ASEC samples from 1988-onward, INCWAGE is derived from OINCWAGE and INCLONGJ. Specifically, OINCWAGE is always a component of INCWAGE. When SRCEARN indicates that INCLONGJ is earned from wage and salary, INCLONGJ is an additional component of INCWAGE. Therefore, for ASEC samples from 1988-onward, you’ll need to use the flag variables for OINCWAGE (e.g., QOINCWAGE) and INCLONGJ (e.g., QINCLONG).

Regarding the discontinuity in the percent of imputed observations: One reasonable explanation for this observation is the change in the method used by the Census Bureau to record information on wage and salary income (see the questionnaire text tab). Prior to 1988, the questionnaire simply asked respondents, “How much did […] earn from this employer before deductions during […]?” From 1988-onward, as discussed above, this information is recorded by asking respondents a number of questions. See the questionnaire text tabs for OINCWAGE and INCLONGJ.